# Hugging Face MCP > Hugging Face MCP gives you access to thousands of pre-trained models, datasets, and interactive demos for NLP, vision, and audio tasks. Instead of navigating dozens of websites, your agent connects directly through this MCP. You can search model architectures by task or author, inspect dataset schemas, run classification jobs, generate text from leading open models, and verify API connectivity all in one place. ## Overview - **Category:** loved-by-devs - **Price:** Free - **Tags:** machine-learning, model-discovery, datasets, nlp, computer-vision, ai-models ## Description This connector lets you interact with the world's largest open-source machine learning hub right from your agent. Need a new model for sentiment analysis? You can find it by searching or browsing curated collections of datasets. Want to test how well an open model generates code snippets? Just run inference, and get results back instantly. It’s all about discovery first. Your agent handles the heavy lifting: finding suitable models, inspecting their metadata, running tests against live demos (Spaces), and finally executing text generation or classification tasks. When you connect this MCP via Vinkius, your AI client treats it like a massive internal resource library—you just ask for what you need, whether that’s checking an account status or listing all available models by author. ## Tools ### list_spaces Searches for interactive ML demo applications (Spaces) to see how others have implemented models. ### check_hf_status Verifies the current operational status and API connectivity to Hugging Face. ### get_account Retrieves your personal account information details from the hub. ### get_dataset Pulls specific metadata and schema details for a given dataset. ### get_model Gets detailed information about a specific model architecture, including usage guidelines. ### get_space Retrieves details for an interactive ML demo application (Space). ### list_collections Lists curated groups of related models, datasets, and Spaces available on the platform. ### list_datasets Searches the hub to find relevant datasets based on keywords or filters. ### list_models_by_author Lists models created by a specific user or organization account. ### list_models_by_task Filters and lists available models based on the machine learning task they perform, sorted by downloads. ### list_models Finds all available models on the Hugging Face Hub using general search criteria. ### run_text_classification Analyzes input text and returns a defined category or label for that text. ### run_inference Executes a model using provided input data and returns the predicted output or classification result. ### run_summarization Sends text to an open model and receives a concise summary of the document's content. ### run_text_generation Generates new, creative, or explanatory text based on a provided prompt using an open model. ## Prompt Examples **Prompt:** ``` Find the top text generation models. ``` **Response:** ``` Top text-generation models: 1) meta-llama/Llama-2-7b (2.1M downloads), 2) mistralai/Mistral-7B (1.8M), 3) google/gemma-7b (950K). Would you like to run inference on any of these? ``` **Prompt:** ``` Generate text with mistralai/Mistral-7B: 'Explain quantum computing in simple terms'. ``` **Response:** ``` Generated (247 tokens): 'Quantum computing uses quantum bits (qubits) that can exist in multiple states simultaneously, unlike classical bits. This enables solving certain problems exponentially faster...' ``` **Prompt:** ``` Search datasets about sentiment analysis. ``` **Response:** ``` Found 15 datasets: 1) 'stanfordnlp/imdb' (25K reviews), 2) 'tweet_eval' (multi-task tweets), 3) 'amazon_reviews_multi' (200K reviews in 6 languages). ``` ## Capabilities ### Search and find model resources Discover models using keywords, filter results by a specific task, or list all available datasets for review. ### Inspect resource details Get detailed information on any given dataset, model architecture, or interactive application (Space). ### Execute machine learning tasks Run live inference jobs to classify text, generate new content, or summarize large documents using open models. ### Manage account information View your profile details and check the current API connection status for troubleshooting. ## Use Cases ### Building a document analysis pipeline A legal tech developer needs to process incoming client contracts. They use their agent to search for appropriate datasets using `list_datasets`, inspect the schema with `get_dataset`, then run classification on sample text using `run_text_classification` to categorize risk levels. ### Quickly prototyping a new feature A startup founder wants to test if an open-source model can write marketing copy. They ask their agent to use `list_models` to find the best text generation tool, then immediately run inference using that tool to generate several headline options. ### Comparing research tools A data scientist needs to compare models for summarization. They ask their agent to use `list_models_by_task` to filter down the options, then use `get_model` on three top contenders to review documentation before running a final `run_summarization` test. ### Debugging an AI integration The DevOps team gets a cryptic API error. They first ask their agent to run `check_hf_status` and then use `get_account` to verify the necessary token scopes, solving the connection issue immediately. ## Benefits - Find the right tools faster. Instead of scrolling through general searches, you can pinpoint exactly what you need by using `list_models_by_task` to see only text generation or classification models. - Test before you commit. You can inspect a model's capabilities and run a live test using `run_inference` against the actual architecture without needing local setup. - Stay updated on performance. Use `check_hf_status` anytime to confirm API connectivity, so your pipeline doesn't fail because of an expired token or outage. - Manage large projects efficiently. You can look through grouped resources using `list_collections`, which is much cleaner than manually tracking dozens of individual datasets and models. - Generate content on demand. Need a few paragraphs explaining quantum computing? Just run the model via `run_text_generation` and get clean, formatted text immediately. ## How It Works The bottom line is you get a single chat interface to manage complex ML workflows without ever leaving your primary application. 1. Your agent first searches or filters resources (like models by task) to identify a potential candidate. 2. You ask the agent to inspect that resource, which pulls detailed metadata and confirms its status for immediate use. 3. The agent executes the required operation—for example, running text generation—and passes the results back directly. ## Frequently Asked Questions **How do I start finding models using list_models by task?** You simply tell your agent you need to find a model for a specific job. The MCP handles the complex filtering, allowing you to see only relevant architectures without manual searching. **Is run_text_generation better than running inference directly?** Both work for generating text, but `run_inference` is a general command that covers all model types. Use `run_text_generation` when your goal is specifically creative or explanatory writing. **Can I check the status using check_hf_status?** Yep, you can. Running `check_hf_status` confirms that all API connections are active and ready to go before you start building a large workflow, saving you time when things break. **What is the difference between list_models and list_models by author?** Use `list_models` for general discovery (e.g., 'show me all sentiment models'). Use `list_models_by_author` when you want to see everything a specific company or researcher has published. **What details does using `get_account` provide about my credentials and permissions?** It returns core account information, including your profile data and organization scopes. This confirms that your agent has the necessary access levels to interact with the Hub's resources. **How does `get_dataset` help me validate if a dataset is usable for my task?** The tool provides the full metadata and schema of the dataset. You can check column names, data types, and required fields immediately before writing any processing code. **If I run too many jobs with `run_inference`, how do I handle potential rate limits?** The MCP handles common API errors, including hitting usage limits. Your agent will receive specific HTTP status codes, allowing you to build proper retry logic into your workflow. **What specifics does `get_model` provide about a model before I decide to run it?** It pulls deep information on the model itself, including its intended use case and specific architecture notes. This helps you verify if the model type matches your exact required task. **Can my AI run inference on Hugging Face models?** Yes. Use `run_inference`, `run_text_generation`, `run_text_classification`, or `run_summarization` to send input to any hosted model and get results instantly. **How do I find the best model for a task?** Use `list_models_by_task` with a pipeline tag like 'text-generation' or 'image-classification'. Results are sorted by downloads so the most popular appear first. **Can I browse datasets and Spaces?** Yes. `list_datasets` and `list_spaces` let you search by keyword, and `get_dataset` / `get_space` return full metadata.